|
Genetic and biochemical approaches have been used to identify and clone regulators of seed development and ABA signaling. The ABI loci of Arabidopsis cloned to date can be divided into genes encoding two major classes of biochemical function: protein phosphatases (ABI1 and ABI2), and transcription factors (ABI3, ABI4 and ABI5). These loci have been well characterized in terms of their roles in marker gene expression and ABA sensitivity of growth, by analysis of mutants and transgenic ectopic expression lines, both singly and in various combinations (Finkelstein and Lynch, 2000; Finkelstein et al., 1998; Parcy and Giraudat, 1997; Parcy et al., 1994; Söderman et al., 2000). These studies have shown that ABI3, ABI4 and ABI5 have similar physiological roles and show some cross-regulation of expression, leading to the suggestion that they act combinatorially. In contrast, the effects of the dominant negative abi1-1 mutation have relatively limited overlap with those of mutations in ABI3, ABI4 or ABI5. Although double mutants combining abi1-1 with any of the transcription factor mutations show significantly enhanced resistance to ABA (Finkelstein, 1994; Finkelstein and Somerville, 1990), suggesting action in separate pathways, the abi1-1 mutation can block the ABA hypersensitivity produced by overexpression of ABI3, consistent with action in the same pathway (Parcy and Giraudat, 1997).
Consequently, ABI1, ABI3, ABI4 and ABI5 were all legitimate candidates for proteins that might be involved in direct physical interactions. We have tested for such interactions among these ABI gene products by yeast two-hybrid assays; as this is a completely heterologous system, any apparent interactions should be dependent on the plant genes included in the assay. Despite the apparent similarities of the genetic interactions involving ABI3, ABI4 and ABI5, only two of the ABI proteins interact in this assay: ABI3 and ABI5. Recent studies have shown that ABI5 interacts synergistically with ABA and co-expressed VP1 to transactivate an Em-GUS reporter in transfected rice protoplasts, providing further support for the hypothesis that these proteins interact in plant cells (Gampala et al., 2001). The failure of ABI1 to interact with any of the transcription factors might reflect either a simple lack of physical interaction between these proteins, or a requirement for a phosphorylated substrate which is not provided by heterologous expression in yeast. The lack of interaction between ABI4 and any of the other ABIs, despite the observed similarities in genetic interactions and physiological effects, does not exclude the possibility of participating in the same regulatory complex, but does show that these factors are probably not in direct contact. We have mapped the ABI3–ABI5 interaction to the B1 domain of ABI3 and a region of ABI5 containing two conserved charged domains, and several possible sites of ser/thr phosphorylation. In addition, ABI5 can form homodimers; this interaction appears to be dependent on a weakly conserved hydrophilic domain as well as the bZIP domain. ABI5/AtDPBF1 homodimer formation has also been demonstrated by in vitro binding to the Dc3 promoter (T. Thomas, Texas A & M University, personal communication).
The functional domains of ABI3 and its homologs, for example, the VP1 proteins of cereals, have been analyzed in terms of their roles in DNA binding and transactivation of target promoters. These studies have shown that the B2 domain is involved in regulation of some late embryogenesis-abundant (LEA) and storage protein genes (Bies-Etheve et al., 1999), and enhances DNA binding of a variety of bZIP factors, but binds DNA only weakly and non-specifically by itself. In contrast, the B3 domain can bind DNA directly (Suzuki et al., 1997), but is not required for ABA-dependent gene regulation (Carson et al., 1997). Both the B2 and B3 domains were recently shown to be necessary for expression of a Brassica napus gene encoding the storage protein napin, apparently by interaction with distinct cis-elements, leading the authors to suggest that B2 tethers ABI3 to a seed-specific ABRE via protein–protein interactions (Ezcurra et al., 2000). Two-hybrid screens for proteins interacting with oat VP1 or Arabidopsis ABI3 using bait constructs containing only the B2 and B3 domains identified several transcription factors, but did not result in isolation of an ABI5 fusion (Jones et al., 2000; Kurup et al., 2000). In contrast, a two-hybrid screen using the B1 and B2 domains of the rice OSVP1 in the bait construct identified TRAB-1 (Hobo et al., 1999), a rice bZIP factor with strong homology to ABI5 (55% similar). Although the interacting domains of TRAB-1 and OSVP1 were not specifically mapped, these results are consistent with the importance of the B1 domain to the interaction.
The functional domains of ABI5 are less well characterized. ABI5 belongs to a subfamily of the bZIP transcription factor family whose members have been identified by mutation (Finkelstein and Lynch, 2000; Lopez-Molina and Chua, 2000), one-hybrid screens with repeats of the ABRE motif (Choi et al., 2000; Uno et al., 2000) or a fragment of the Dc3 promoter (Kim et al., 1997); or a two-hybrid screen with OSVP1 as ‘bait’ (Hobo et al., 1999). Within this subfamily, expression patterns vary in terms of relative abundance in seed versus vegetative tissues or induction by salt, drought or ABA, but all appear correlated with ABA response. To date, the only member of the family for which mutants have been identified is ABI5. The most severe available mutations result in truncation at aa 361 (abi5-1) (Finkelstein and Lynch, 2000) or after only 110 amino acids of the wild-type coding sequence (abi5-4) (Lopez-Molina and Chua, 2000); the former includes all but the bZIP domain, while the latter is lacking all but the first conserved charged domain. However, both mutant lines show similar, limited ABA resistance and comparable effects on seed gene expression, suggesting that both are effectively null mutations. Given that both lack the bZIP domain required for dimerization and DNA binding, this result is neither surprising nor informative regarding other potentially important domains. In contrast, deletion analyses of the sunflower DPBF-1 protein have shown that in vitro DNA binding is retained in constructs lacking one or all three of the charged domains in the amino half of the protein (Kim et al., 1997). This implies that the region of DPBF-1 corresponding to amino acids 1–248 of ABI5 is not required for homodimer formation. Although sunflower DPBF-1 shows the strongest overall homology to ABI5 of all sequences currently in the database, the Arabidopsis ABF/AREB proteins share with ABI5 a larger conserved region at the amino-terminal end. In our studies, this conserved region appears to function as a transcription-activation domain. Although a truncation including only the most conserved part of this domain (aa 64–122) still functions as an activator, it is attenuated approximately twofold relative to the construct including amino acids 9–122. Recently, all three of the conserved domains in the amino-terminal halves of AREB1 and AREB2 were shown to undergo ABA-dependent phosphorylation in an in-gel kinase assay (Uno et al., 2000). Furthermore, transactivation of reporter genes in protoplasts required the presence of both an AREB and ABA, but could be blocked by protein kinase inhibitors, implying that ABA-dependent phosphorylation is essential for activity of the AREBs. However, in vivo phosphorylation of the AREBs has not been assayed directly.
Taken together with our two-hybrid interaction results, these studies suggest that the transcription-activation domain of ABI5 is located in the most amino-terminal conserved domain, but is normally inactive until phosphorylation induces a conformational change that exposes the domain to potential interaction partners. In our truncation studies, removal of surrounding domains by deletion of their coding regions might produce the same effect as that suggested for phosphorylation: exposure and resulting transcription-activation function in yeast. In contrast, the second and third conserved domains appear more significant in interacting with ABI3, but have no intrinsic activator function of their own. We do not know whether this interaction depends on phosphorylation of the BD-ABI5 truncations in yeast, but they are certainly not exhibiting ABA-dependent phosphorylation in this system. Analysis of the predicted protein structure for ABI5 shows that the three conserved regions in the amino half have a relatively low probability of being exposed on the surface compared with adjacent domains in the primary structure. This is consistent with a requirement for a phosphorylation-induced conformational change to expose the transcription activation and ABI3-interacting domains.
The studies described above all use the GAL4 DNA-binding domain to target transcription activation due to any interacting proteins to a GAL4-responsive promoter. To determine whether any of the ABI-transcription factors could interact directly with a known ABI-responsive plant promoter in a completely heterologous system, we tested their ability to activate an AtEm6::lacZ fusion in yeast. The promoter fragment used contains two consensus binding sites (ACACNNG) for the sunflower homologs of ABI5 (Kim et al., 1997) within ≈230 bp of the transcription start site, and it is likely that the observed activation by ABI5 reflects binding to these sites. In contrast, intact ABI3 (or VP1) has never been shown to bind directly to DNA, so the failure of ABI3 to transactivate the reporter gene is not surprising. Similarly, although ABI4 belongs to the EREBP/DREB/CBF subfamily of the APETALA2 domain family of transcription factors, most of which were identified by binding to similar DNA sequences, a binding site for ABI4 has not yet been identified. The failure of ABI4 to transactivate the AtEm6::lacZ reporter suggests that an ABI4 binding site either is not contained within this 1.3 kb fragment, or is too far from the transcription start site to efficiently recruit RNA polymerase to initiate transcription of this gene.
|
